remove selfmon (#86)

Author: DuodenumL

Makefile

@@ -25,7 +25,6 @@ unit-test:
 	go test -race -count=1 -timeout 240s -cover ./logs/... \
 	./manager/node/... \
 	./manager/workload/... \
-	./selfmon/... \
 	./types/... \
 	./utils/...
 

README.md

@@ -68,4 +68,4 @@ Make sure you can clone code by ssh protocol because libgit2 ask for it. So you
 <cli_execute_path> container deploy -pod <pod_name> --entry agent --network <network_name> --deploy-method fill --image <projecteru2/agent>|<your_own_image> --count 1 --file <agent_config_yaml>:/etc/eru/agent.yaml [--cpu 0.3 | --mem 1024000000] http://bit.ly/EruAgent
 ```
 
-Now you will find agent was started in each node, and monitor containers status include itself.
+Now you will find agent was started in each node, and monitor containers status include itself.

agent.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"context"
 	"fmt"
 	"math/rand"
 	"os"
@@ -12,7 +13,6 @@ import (
 	"github.com/projecteru2/agent/api"
 	"github.com/projecteru2/agent/manager/node"
 	"github.com/projecteru2/agent/manager/workload"
-	"github.com/projecteru2/agent/selfmon"
 	"github.com/projecteru2/agent/types"
 	"github.com/projecteru2/agent/utils"
 	"github.com/projecteru2/agent/version"
@@ -56,17 +56,12 @@ func serve(c *cli.Context) error {
 	utils.WritePid(config.PidFile)
 	defer os.Remove(config.PidFile)
 
-	if c.Bool("selfmon") {
-		mon, err := selfmon.New(c.Context, config)
-		if err != nil {
-			return err
-		}
-		return mon.Run(c.Context)
-	}
-
-	ctx, cancel := signal.NotifyContext(c.Context, syscall.SIGINT, syscall.SIGTERM, syscall.SIGQUIT)
+	ctx, cancel := context.WithCancel(c.Context)
 	defer cancel()
 
+	signalChan := make(chan os.Signal, 1)
+	signal.Notify(signalChan, syscall.SIGINT, syscall.SIGTERM, syscall.SIGQUIT, syscall.SIGUSR1)
+
 	errChan := make(chan error, 2)
 	defer close(errChan)
 
@@ -103,10 +98,18 @@ func serve(c *cli.Context) error {
 	go func() {
 		select {
 		case <-ctx.Done():
-			log.Info("[agent] Agent caught system signal, exiting")
+			log.Info("[agent] Agent exiting")
 		case <-errChan:
 			log.Info("[agent] got err, exiting")
 			cancel()
+		case sig := <-signalChan:
+			log.Infof("[agent] Agent caught system signal %v", sig)
+			if sig != syscall.SIGUSR1 {
+				if err := nodeManager.Exit(); err != nil {
+					log.Errorf("[agent] node manager exits with err: %v", err)
+				}
+			}
+			cancel()
 		}
 	}()
 
@@ -234,17 +237,6 @@ func main() {
 				Usage:   "change hostname",
 				EnvVars: []string{"ERU_HOSTNAME"},
 			},
-			&cli.BoolFlag{
-				Name:  "selfmon",
-				Value: false,
-				Usage: "run this agent as a selfmon daemon",
-			},
-			&cli.StringFlag{
-				Name:    "kv",
-				Value:   "",
-				Usage:   "kv type",
-				EnvVars: []string{"ERU_AGENT_KV"},
-			},
 			&cli.BoolFlag{
 				Name:  "check-only-mine",
 				Value: false,

agent.yaml.sample

@@ -12,10 +12,6 @@ store: grpc
 # This option is not required as the default value is "docker".
 runtime: docker
 
-# kv defines the type of kv store.
-# This option is not required as the default value is "etcd".
-kv: etcd
-
 # core defines the address of eru-core component.
 # This option is not required as the default value is "127.0.0.1:5001".
 core:
@@ -24,10 +20,8 @@ core:
 
 # heartbeat_interval defines the interval for eru-agent to
 # report health status of the node to eru-core.
-# This option is not required, and is only useful when enabling
-# selfmon mode.
 # If you don't want eru-agent to report this status, set it to 0.
-# The default value of this option is 0.
+# The default value of this option is 60.
 heartbeat_interval: 120
 
 # auth defines the authentication values for eru-core.
@@ -117,41 +111,13 @@ log:
 #
 # healthcheck.cache_ttl defines how long will eru-agent cache an unchanged status locally.
 # This is only used when selfmon mode is switched on. The default value is 300 (in seconds).
-#
-# healthcheck.enable_selfmon defines whether selfmon is switched on.
-# This should be true if there's at least one eru-agent is in mode selfmon.
-# When this is true, healthcheck.enable_selfmon and heartbeat_interval is meaningless.
 healthcheck:
   interval: 120
   timeout: 10
   cache_ttl: 300
-  enable_selfmon: false
 
 # global_connection_timeout defines the timeout for eru-agent other than healthcheck.
 # E.g. the timeout for reporting action of eru-agent, or the timeout for eru-agent to
 # connect to docker.
 # The default value is "5s", note that "s" in the end.
-global_connection_timeout: 15s
-
-# ha_keepalive_interval defines the time interval for sending heartbeat
-# when selfmon maintains its own active state.
-# The default value is "16s", note that "s" in the end.
-ha_keepalive_interval: 16s
-
-# etcd defines the etcd configuration.
-# This option is required and has no default value.
-# If you don't plan to run this eru-agent in selfmon mode,
-# you can give a mocked value e.g. 127.0.0.1:1111,
-# this value won't be used to connect, it's only to pass
-# the validation of this option (it's tricky).
-# Will plan to improve this in next release.
-#
-# etcd.machines defines the addresses of etcd machines.
-#
-# etcd.prefix defines the prefix for eru-agents in selfmon mode.
-# This prefix should be the same for all eru-agents in selfmon mode,
-# and also distinguished for different ERU clusters.
-etcd:
-  machines:
-    - 127.0.0.1:2379
-  prefix: /agent-selfmon
+global_connection_timeout: 15s

go.mod

@@ -12,7 +12,7 @@ require (
 	github.com/go-ole/go-ole v0.0.0-20180213002836-a1ec82a652eb // indirect
 	github.com/jinzhu/configor v1.2.1
 	github.com/patrickmn/go-cache v2.1.0+incompatible
-	github.com/pkg/errors v0.9.1
+	github.com/pkg/errors v0.9.1 // indirect
 	github.com/projecteru2/core v0.0.0-20211021040158-0be8dbadbc55
 	github.com/projecteru2/libyavirt v0.0.0-20211014062234-66e6f24ab6d1
 	github.com/prometheus/client_golang v1.11.0

manager/node/manager.go

@@ -78,22 +78,24 @@ func (m *Manager) Run(ctx context.Context) error {
 	log.Info("[NodeManager] start node status heartbeat")
 	go m.heartbeat(ctx)
 
-	// wait for signal
 	<-ctx.Done()
-	return m.exit()
+	log.Info("[NodeManager] exiting")
+	return nil
 }
 
-func (m *Manager) exit() error {
+// Exit .
+func (m *Manager) Exit() error {
 	log.Info("[NodeManager] exiting")
-	log.Infof("[NodeManager] mark node %s as down", m.config.HostName)
+	log.Infof("[NodeManager] remove node status of %s", m.config.HostName)
 
 	// ctx is now canceled. use a new context.
 	var err error
 	utils.WithTimeout(context.TODO(), m.config.GlobalConnectionTimeout, func(ctx context.Context) {
-		err = m.store.SetNode(ctx, m.config.HostName, false)
+		// remove node status
+		err = m.store.SetNodeStatus(ctx, -1)
 	})
 	if err != nil {
-		log.Errorf("[NodeManager] failed to mark the node %s as down, err: %s", m.config.HostName, err)
+		log.Errorf("[NodeManager] failed to remove node status of %v, err: %s", m.config.HostName, err)
 		return err
 	}
 	return nil

manager/node/manager_test.go

@@ -23,10 +23,9 @@ func newMockNodeManager(t *testing.T) *Manager {
 			Stdout: true,
 		},
 		HealthCheck: types.HealthCheckConfig{
-			Interval:      10,
-			Timeout:       5,
-			CacheTTL:      300,
-			EnableSelfmon: true,
+			Interval: 10,
+			Timeout:  5,
+			CacheTTL: 300,
 		},
 		GlobalConnectionTimeout: 5 * time.Second,
 	}

manager/workload/manager_test.go

@@ -23,10 +23,9 @@ func newMockWorkloadManager(t *testing.T) *Manager {
 			Stdout: true,
 		},
 		HealthCheck: types.HealthCheckConfig{
-			Interval:      10,
-			Timeout:       5,
-			CacheTTL:      300,
-			EnableSelfmon: true,
+			Interval: 10,
+			Timeout:  5,
+			CacheTTL: 300,
 		},
 		GlobalConnectionTimeout: 5 * time.Second,
 	}